Home / Articles / Volume 7 (2010) / To quote or not to quote: Setting the context for computer-mediated dialogues
Document Actions



Electronic mail (email) has increasingly become recognized as a medium for informal and interactive discourse. It is often characterized as conversational, although the conditions for a dialogue deviate from those for face-to-face interaction in many respects (Severinson Eklundh, 1986). The differences are not only due to the fact that email is a written medium, but also to the time delay between successive contributions and the lack of immediate feedback in email discourse. Furthermore, both email and other, related forms of computer-mediated communication (CMC) are characterized by the possibility of an individual being simultaneously involved in several ongoing exchanges (Herring, 1999).

These properties of the computer medium may pose problems for participants, particularly in creating coherence in busy network communication. Placing a message in its proper discourse context is crucial both for the sender, who must indicate clearly how the message relates to the previous discourse, and for the receiver, who has to use available cues to establish its conversational relevance.

Many CMC systems support the simple context-linking mechanism of quoting, i.e., repeating (including) a message or part of it as a part of the reply. For the receiver, quoting serves to situate the response in a discourse context, e.g., in a situation with multiple ongoing exchanges, and thus facilitates the perception of an extended conversation as coherent. For the sender, it facilitates composition by allowing direct response without having to paraphrase the original message.

These properties of quoting were frequently referred to as advantages in a survey of email communication among readers of Usenet newsgroups1 conducted by Severinson Eklundh and Macdonald (1994). The results of the survey, which was returned by 250 individuals from different countries, confirmed that quoting was widespread at the time and was viewed as an essential part of communication on busy electronic networks.

At the same time, there are potential problems associated with quoting as a conversational strategy. Including a previous message makes the reply longer and may cause less than ideal reading conditions on the computer screen. Moreover, verbatim repetition of an interlocutor's utterance may be perceived as a violation of everyday norms of conversation. Many respondents in the Usenet survey characterized quoting as impersonal or impolite, and they reported that they avoided it in personal email communication. The appropriateness of quoting was found to depend on a number of different conditions in the communicative situation, such as the degree of formality of the exchange, the number of receivers of the message, and so on.

As a conversational strategy, quoting can be seen as embodying a pragmatic conflict. On the one hand, it helps place a message in context, thereby providing clarity about the state of an interaction and establishing coherence. But on the other hand, it adds significantly to the length and redundancy of a message, potentially making it difficult to read. As will be shown below, this can be seen as an opposition between different conversational maxims (Grice, 1975). Against this background, it is of great interest to explore how participants use quoting in actual situations, and how their strategies vary across different varieties and genres of CMC.

The need for explicit contextualization of a message can be viewed as particularly important when many conversations are going on simultaneously, which is typically the case in large-group discussions. This was supported by the Usenet survey, in which participants reported quoting to be particularly frequent in newsgroup communication. However, simultaneous dialogue threads also occur in other forms of CMC, such as private or small-group email exchanges, and they are sometimes described as typical of the asynchronous computer medium (Black, Levin, Mehan, & Quinn, 1983).

In this article I analyze quoting in computer-mediated conversations from a pragmatic perspective, and examine closely its role in two communicative contexts: private email and Usenet newsgroups. The following issues are in focus:

•How does the use of quoting for contextualization differ between newsgroup communication and private email? In particular, do participants in Usenet discussions quote more frequently than email authors?
• To what extent do participants in these two settings strive for economy in quoting, so that only relevant portions of text are included?
• What are the roles of quoting for the conversational character and the structure of computer-mediated exchanges?

By means of descriptive analysis of data collected from these two CMC settings, I highlight the variability of the use of quoting across various contexts and its different functions in computer-mediated dialogue. In the subsequent discussion, I argue that technical properties of the mail system and factors in the social context jointly affect the quoting strategies of CMC participants.

Quoting as a Pragmatic Choice

Quoting can be seen as a pragmatic and interactional resource, which by convention can be employed to contextualize a response in CMC. Electronic mail systems give various forms of support for this convention by offering commands for including and editing the text of a previous message in a message being composed.

Pragmatically, the quoted text has a discourse-deictic relationship to the message from which it is taken (cf. Levinson, 1983). It functions as a substitute for a previous message which is currently being addressed in a new conversational move. This phenomenon is distinct from traditional forms of quotation in writing, defined, e.g., as "repeating as by way of authority or illustration" or "bringing forward as evidence or support" (Webster's, 1993). Quoting is also different from the various forms of repetition that abound in spoken discourse.2 In fact, repetition of an entire utterance merely for establishing reference is a strategy unknown in other forms of interactive communication and thus has to be learned by users of electronic media.

In this article, I will only be concerned with the use of quotation, unique to computer-mediated communication, which serves to link together messages in a conversational thread. Quotation is also used for other purposes in CMC (see, e.g., Hodsdon-Champeon, this issue). In fact, one could argue that the term "quoting" itself, although widespread, is too broad for the phenomenon examined here. For the purpose of clarity, I sometimes use the longer form "contextual quotation" to refer to the mechanism of including a message or part of one in the reply to a message, and when no misunderstanding can arise, I use "quotation" as the short form for this phenomenon.3

Technically, contextual quotation is accomplished by the sender attaching a previous message or part of it to the message currently being composed. This is usually done by a command or a setting in the mail system, which either inserts the whole message or allows the user to specify the part to be inserted. The response itself can be written either before or after the inserted (quoted) text, which is automatically marked or indented by the system to separate it from the new text (see Example 1). Usually, a special quotation sign, such as ">" or ":" is inserted before each line of quoted text, but other conventions for marking the quote are also used.

Although the use of quoting is usually optional, the design of the mail system can in various ways reflect what is considered to be normal behavior by making certain choices particularly simple. For example, some systems include the whole message by default in the reply, forcing the user to delete explicitly any text that he or she does not want to quote, whereas others allow the user to select a portion of text to be included.

As a part of the quoting facility, it is possible in many systems to attach a pointer line in the beginning of the quote, stating who is the sender of the message being quoted (e.g., "In message xyz, Joe@abc.com wrote:"). This helps the reader follow how a conversation develops in a group with multiple interlocutors and several ongoing branches of a single discourse topic (see Example 1).

By using a text editor, the author can choose to delete parts of the inserted message that are considered less relevant to the response. One may also intersperse one's comments within the quoted passage, so that different parts of a message are followed directly by their successive responses in a topic-response fashion. The respondents of the email survey in Severinson Eklundh and Macdonald (1994) reported that they used the interspersed strategy frequently, and many of them commented that it gives the reply a conversational character. Example 1 illustrates this strategy of responding.4 In this interspersed quotation-response, the entire text of the antecedent message has been included in the quote except for an introductory line reading "I just read the message from John."

Example 1

This form of reply, containing "multiple conversational turns" in the same message, has been characterized as a prototypical pattern in Usenet newsgroup discourse by several authors (e.g., Hodsdon-Champeon, this issue; Reed, 2001). Hodsdon-Champeon examined quoting as one of several forms of intertextuality within a newsgroup discussion on racial discrimination. An interesting finding of her study is that direct quotations were often used to mark a distance from the view expressed by an interlocutor. Persons who quoted text from a previous message usually displayed a negative stance towards the views expressed in the quote.

In both email and newsgroup discourse, a quoted message may itself contain one or more quotations from previous messages. Extended conversations may emerge, where each message includes quoted parts from the previous one. The last message will function as a summary of a whole conversation, and may be sent or stored in its entirety, as an object of its own. This aspect of quoting in email as a “history mechanism” is discussed in Duchenaut and Bellotti (2003).

Because quoting adds substantially to the amount of text in a message and may distract from the actual response, there is a risk that by using it, participants will violate the conversational maxims that characterize rational, cooperative communication (Grice, 1975; Levinson, 1983). For example, by quoting a lengthy message while addressing only part of it in the response, the contribution will contain more information than required (cf. the maxim of Quantity), contain material not relevant to the response (maxim of Relevance), and be less brief than it should be, or even obscure (maxim of Manner).

In fact, there have evolved special rules of conduct on electronic networks, especially in newsgroup communication, to the effect that messages should be as brief and to the point as possible. A summary of such rules is given in the "Taxonomy of Reproachable Conduct on Usenet" compiled by McLaughlin, Osborne & Smith (1995). The taxonomy was derived from the analysis of an extensive corpus of corrective episodes in Usenet newsgroups, i.e., conversations where a message or posting of a particular type was followed by a response in which the conduct in the posting was reproached by another participant. The taxonomy lists the case of "quoting material longer than original in follow-on" as an example of "Bandwidth waste" (explained as "consuming more than one's fair share of network resources," p. 98). At the same time, the convention of quoting itself is established as a norm in the taxonomy, by listing "Failing to include text to which one refers in a follow-on" as a case of "Violation of networkwide conventions" (p. 97).

Thus, although the strategy of quoting is enforced by newsgroup conventions, participants are expected to limit their quotations to ensure brevity and relevance. In a study of politeness phenomena in CMC by Herring (1994), "quoting the entire previous message" was included among violations of negative politeness in the sense of Brown and Levinson (1987).5 The results of a survey among users of eight discussion lists showed that this strategy was among the most negatively ranked of net behaviors, only exceeded by the posting of racist and sexist messages.

Quoting potentially entails a conflict in applying the conversational maxims of Quantity and Relevance. On the one hand, it serves to contextualize the response, thereby making it more relevant. On the other hand, the quotation makes the message longer and sometimes difficult to read. A crucial question is how users choose to deal with this conflict in actual situations, and in particular how features of the context and systems factors may influence their strategies of quoting.

Although quoting is an established convention in many forms of electronic group communication, there is great variation in its use in different CMC contexts. This was confirmed in the corpus of data from ProjectH, a CMC project conducted in collaboration among an international group of researchers (Sudweeks & Rafaeli, 1994). A large body of messages was collected and coded with respect to a number of content and structural variables. The percentage of messages containing quotations in this corpus ranged from 1% to 100% across the different discussion groups studied.

Below, I examine the quoting strategies used in a sample of private email communication and a sample of newsgroup communication, respectively. The goal is to clarify the extent to which quoting is used in those two contexts; how participants select the portion of quoted text in relation to the text in the reply, and how the quoting mechanism affects the conversational character of the exchange.

The study uses the following sources of data:

1. The personal mailbox during one month of a female academic who uses email for communication at work. These data serve to illustrate how quoting can be used to provide a discourse context within private email communication, both locally and across larger distances.
2. Discussions held during one month within the Usenet newsgroup comp.human-factors. These data are analysed to show how multi-party conversational threads develop in an environment where quoting is a strongly enforced and conventionalized discourse mechanism.

Exploring Quotation Patterns in Email and Newsgroup Communication

Electronic mail and newsgroup discussions are established and widespread forms of CMC, each with its own characteristic properties. Whereas messages in the newsgroup are directed to a large number of people (in fact, the writer usually does not know exactly how many people will read the message), email messages typically have only one or a few recipients. Studies of private email have been quite rare in research about conversational aspects of CMC, presumably because of problems in acquiring data and guaranteeing the integrity of the informants. There is therefore general interest in studying email and its conversational properties compared with the more commonly studied group discussions.

The two types of data also exhibit differences with respect to the nature of the discourse produced. The purpose of the newsgroup is to provide a forum for discussion and information exchange among people with similar interests. The messages are often argumentative and sometimes raise more than one issue. In contrast, email messages typically deal with only one topic or issue at a time, and have a more transactional character. The conversational threads in newsgroups also tend to be longer than in email – in fact, the prototypical newsgroup message can be said to be a reply, since each new initiative typically gives rise to more than one response. Due to these differences, the roles of quotation as a reply mechanism can be expected to differ in the two types of data, although its basic function in both cases is to provide a context for the reply.

Yet another difference is that the email conversations analyzed here were mainly performed in Swedish, whereas Usenet newsgroups have an international user community, often dominated by Americans, and discussions are usually carried out in English. However, most Swedish academics are connected to the international networks, and the style of their electronic communication is heavily influenced by this international communication.

The many differences in the nature and purpose of communication in these settings mean that it would not be meaningful to make a rigorous statistical comparison between the two sets of data. Also, the material collected is rather limited, which would make such a comparison problematic. Instead I have chosen to make a careful descriptive analysis of the quotation patterns found in these two settings and contrast them across a number of dimensions. Subsequently, I use the results to discuss the impact of the context as well as the communication system on the participants' quoting strategies.

The Data

The Mailbox of an Academic Email User

Permission was granted to collect one month's worth of email from the electronic mailbox of a female academic (henceforth: A) working in a human-computer research laboratory in Sweden. She was experienced in using computers without being technically trained and had used email for nine years at the time of the data collection (fall 1996).6

A's email communication concerned mainly work issues, involving close and distant colleagues as well as students. Some of it also dealt with issues around her Ph.D. thesis defense in psychology, which was due shortly after the data were collected. She subscribed to some international mailing lists which generated mainly announcements of a professional character. Except for these, A's email contacts were almost exclusively Swedish during the period of data collection.

In total, the collected data comprised 364 messages, of which 116 were written by A herself and 248 were messages directed to her. Since the subject of investigation was how quoting was used to contextualize a response, only the messages that were part of a coherent thread (dialogue) were subjected to analysis. This meant that a number of single messages, most of them from professional mailing lists, were excluded from consideration.

Usenet Discussions on Human-factors Issues

During one month in the autumn of 1996, a body of messages was extracted from the Usenet newsgroup comp.human-factors. The topics of discussion in this group concern mainly human-computer interaction and interface issues. The messages were organized into threads according to the automatic linking of the system, and the ones that were not followed by a reply (about 100) were excluded from analysis.7 This resulted in a corpus of 23 separate discussions comprising 2 to 28 messages each.

In both sets of data, the focus of study was the strategies used to link messages with subsequent replies. All messages whose subject line started with Re: were analyzed, both according to a number of quantitative characteristics related to quoting, and with respect to their relationships to the surrounding dialogue. The analysis included a range of aspects such as the amount, nature, and placement of quoted text, as well as the use of pointers and epistolary conventions (e.g., greetings and signatures) together with quoting.

Quantitative Account of Quoting Patterns

In this section I present a number of quantitative characteristics of the data in order to provide a picture of how quoting was used in the two communicative contexts and the relationships of quoting to other aspects of a reply. Table 1 gives a summary of these characteristics of the data.8

Frequency of Quoting

The frequency of quoting can be expected to reflect its general importance as a contextualizing mechanism in the two settings studied. In particular, it is of interest to examine if quoting is more widely used in the newsgroup data.

In Table 1, I have separated messages from A (A-sent) from the messages directed to her (A-received). The reason is the bias when one single author dominates a corpus. At some points, however, I also present the data together. The body of reply messages in A-received contained 72 items. In all, 38 of these messages (53%) contained a contextual quotation (i.e., a quotation used to contextualize the response). In addition, three messages contained an attachment message that was not characterized as contextual. A's own messages displayed an even higher quoting frequency: Out of 50 reply messages she sent during one month, 36 (72%) contained a contextual quotation.

There were 28 different persons who had written a reply message to A during the month of the data collection, 18 of whom used quoting at least once. Of the 16 persons who wrote two replies or more, three quoted each time, 11 quoted in certain cases but not in others, and two persons did not quote in any of their messages in the data. This is in line with the survey results discussed earlier, to the effect that for most email users, quoting is a choice that is dependent on the situation.

The newsgroup data contained 117 reply messages in all (i.e., the Subject line started with Re:). Among them, 84% contained a contextual quotation from a previous message in the thread. This quoting frequency is much higher than in the private email corpus. In fact, 68 of the 76 contributing newsgroup authors used contextual quoting at least once. Their pattern was also more consistent than the email authors: Of the 20 authors who provided two replies or more in the data, 15 quoted each time, four quoted at least once and avoided quoting at least once, whereas only one avoided quoting altogether.

Table 1. Summary of quantitative aspects of quoting in the data

Selection of Material to be Quoted

According to the conversational maxims, a CMC user ought to select carefully what to include in a quotation in order to provide clarity and avoid sending unnecessary information. Therefore, it is of interest to examine to what extent participants include an entire message, and to what extent they select a particular part to quote and respond to in their reply.

In the A-received part of the email data, a relatively large proportion of the replies with a quotation (21 of 38, or 55%) included the whole quoted message. (Several of them had removed greetings and signatures from the quote, however; see further below). This frequency is higher than expected, considering the norms against "wasting bandwidth" in network communication. Clearly, these users are not influenced by network etiquette to the same extent as the respondents in the Usenet survey, almost 90% of whom reported that they seldom or never quote the whole text of a message. Only eight messages in A-received included a selected part of the message, and in the remaining nine cases, the antecedent message from which the quote was taken was missing from the corpus, so it was impossible to tell how much of it was quoted.

A's own strategy was quite consistent. When she used quotation, she always included the whole message, including greetings and signatures. There were 31 such cases in A-sent, and five additional cases had a missing antecedent. In total, 70.3% of the quoted email replies included the whole message, whereas only 11% contained a selected part.

In contrast, the participants in the newsgroup were much more selective in their use of quotation. Only 19 of the 98 replies with a quotation (19.4%) included the whole message, and most of those had removed all greetings and signatures from the quote. The majority, 59 messages, contained a selected part of the included message. In the remainder (19 messages), the antecedent message was missing from the corpus.

An interesting issue is how the writer selects the part of the text to be quoted. Great variation is to be expected, since anything relevant to the response can be quoted. In terms of linguistic units, however, there seems to be a tendency to quote elicitative parts of a message as the primary choice. For example, if a message starts with a background passage (a grounding or "premonition" to a question, cf. Severinson Eklundh, 1986), and ends with a question, the quotation often includes precisely the last, questioning part. Example 2 illustrates this case.

Original message (an attached quotation is removed):

In the reply message, only the question is quoted:

Example 2

Often, quoting does not follow syntactic boundaries. This is explained by the fact that mail programs often make it easier to edit the message by lines rather than linguistic units; see Example 3.

Original message:

Reply from one of the receivers:

Example 3

Length of Responses and Quotations

Since long quotations potentially imply a violation of network conventions, it is of interest to examine how the presence of a quotation actually affects the length of messages across different communicative contexts. In the data collected, the number of lines of text in the quotation and response parts of a message were counted exclusive of all headers, greetings, and pointers; the latter were analyzed separately.

The longest reply messages were produced in the newsgroup setting (mean length 17 lines compared to 11.6 lines in A's email data). In spite of this, the quotation parts of the texts were shorter in the newsgroup (mean 6.2 vs. 8.3 lines in A's mailbox). This confirms that the newsgroup participants are more selective in quoting from other messages. On average, their quotes were about half the length of the original response parts of messages, whereas email authors quoted roughly the same amount of text as they wrote in their responses.

A's own replies were generally longer than the ones she received (13.3 lines vs. 10.5 lines in A-received). She also included somewhat longer texts in her quotations than the individuals she communicated with (9 lines vs. 7.6 lines in A-received).

A closer analysis reveals an interesting difference between the two sets of data in the relationship between quotations and response parts of messages. In the newsgroup messages containing a quote, the length of the response part increases with the length of the quote (see Figure 1). In contrast, the email data, both A-received and A-sent, show a decline in response length among the messages with the longest quotes. One interpretation of this is that newsgroup participants quote only those parts of a message that they wish to comment on, and therefore, a longer quote implies a more comprehensive response. Email authors in A's data, in contrast, often quote the whole message regardless of the length of the comment they want to make. Another relevant circumstance is that messages tend to get longer during the course of a dialogue because of embedded quotations. However, this still does not explain why the longest quotes in the email data are coupled with shorter responses than the short quotes. One might speculate that there is a maximum ideal length of a message, which may keep email authors from writing excessive response parts together with contextual quotations. More research is needed to explore these relationships further in different CMC settings.

Figure 1. Length diagrams of replies, and their quotation and response parts, in the three types of data (newsgroup, received mail and sent mail)

Black = quotation part; white = response part of the reply

1. Mean length of replies without a quotation.

2. Mean length of the half of replies with shorter quotations.

3. Mean length of the half of replies with longer quotations.

The Position of the Quoted Passage in Relation to the Text

As mentioned before, there are different ways of placing the quoted text when inserting it into the message being composed. It may be inserted either at the top of the message, with the response following after it, or at the bottom of the message, after the response. Intuitively, placement at the top lends more of a character of dialogue to the entire message, in that a "question" or topic precedes the corresponding response. This is also the most natural way of interspersing topics and responses, giving the entire message a dialogue-like character (see, e.g., Example 1).

In contrast, placing the included message at the bottom makes the quotation more like a separate attachment to the message. Hypothetically, this might enforce a different model of email discourse, in which the quotation is a separate text delivered together with the response.

An email system may facilitate one of these alternatives to a greater extent than the other. For example, in the system used by A at the time of this study (Pine) the quotation was placed below the cursor, which made it natural to write the response at the top of the message, at least if the quotation was not to be edited but included in its entirety.

Of the 38 messages with a quotation in A-received, 22 had the quote below the response, 13 had it above the response and six messages displayed the interspersed pattern. A's own messages were completely homogeneous: All of the 36 quotations were placed below the response.

This is in sharp contrast with the newsgroup data, where the quotation is almost always placed above the response. Only eight of the 98 reply messages with a quote contained a quoted passage below the response. Eighteen had quotes and responses interspersed, whereas the majority (72, i.e., 73%) of the responses with quotes had the quoted passage at the top. This appears to be a predominant strategy in newsgroup discourse. It is also in line with the preferences expressed by participants of the Usenet survey, where only 2% of the 250 participants reported that they always placed the quote below the text.

Most participants in both contexts are consistent in their manner of placing the quotation. In the newsgroup data, however, eight individuals alternate between placing the quote above the response and interspersed with it. In A's email data, three individuals alternate between placement below and interspersed, and two individuals alternate between placement below and above the response.

The Character of the Response when Quoting is Used

An interesting issue with respect to quoting is how it affects the conversational character of the communication. In the Usenet email survey, we asked this question explicitly. A majority of respondents commented that it increases the conversational character of an email exchange. Among the reasons mentioned were the fact that quoting allows direct response, and that the interspersing of topics and responses lends the whole exchange a dialogue-like character.

One way of assessing this similarity in actual data is to examine how the response connects linguistically to the quoted text. A typically conversational way of connecting to a previous turn in a reply is to start with one-word affirmative or negative responses (such as "Yes" or "No"; see Example 1) or with an exclamation (such as "Great!", "Thanks!", "Wow" etc.). These kinds of elements, called interjections here, often appear as independent utterances in spoken discourse (McLaughlin, 1984, p. 95). Another conversational strategy is to use anaphoric elements in the response that refer to elements in the quoted passage (for example, to use the pronoun "it" to refer to an object mentioned in the quoted passage; see Example 2). One might hypothesize that such direct, elliptic answering strategies would occur more frequently if participants experience the communication as conversational, whereas a more "written" way of expression would use paraphrase in these cases (cf. Korenman & Wyatt, 1996 9).

The analysis shows that interjections, along with many other forms of elliptical expressions, are common as part of responses in both sets of data (see Table 1). Notably, more than 40% of the newsgroup reply messages contained an anaphoric reference to its antecedent message or quotation part. In the email corpus, the use of interjections and anaphora is more common in replies with quotes than in others (although A's own messages contain very few interjections throughout). It is difficult to compare the frequencies of interjections across the two types of data, since the possibility of such elements in a response depends to a significant extent on the form of the original message.

Embedded Quotations

When a message is quoted to provide a context, there may already be a quoted passage in the message being included. Including this part as well makes the entire reply longer and potentially difficult to read due to double markers and indentations. On the other hand, it can be argued that the quotations give a picture of the structure of the dialogue that help to create a context. Considering the maxims of Quantity, Relevance and Manner, it is of interest to examine to what extent CMC users do include previous quotes in a quotation.

In all, around one-fourth of the quoted messages in comp.human-factors, and one-fifth of the quoted messages in A's mailbox data, contain embedded quotations. In general, the embedded quotations in A's mailbox are inserted as is, which often makes the whole message significantly longer, whereas newsgroup authors often edit the embedded quotes to contain only the parts they want to address. In Example 4, both of the quotes have been edited, the second one cut off in the middle of a sentence.

Example 4

Example 5

When repeated quotation is used and the participants follow different practices in terms of where they place the included text, the result can be quite confusing. This is illustrated in Example 5 from A's email data. The sender quotes a previous message and inserts a brief comment. However, the quoted message also contains an embedded quotation which is left at the end. The result is a long and complex message in which the three-word response is difficult to locate. Another consequence of the embedded quotation is that the reply ends with two signatures by the same individual, one of them within quotation signs.

Combining Quotation with Epistolary Conventions

Greetings and signatures are characteristics of a message that are less compatible with the conversational aspect of email. They lend a formal character to the message reminiscent of traditional letters. They also constitute parts that should not be included in a quote, according to the maxim of Relevance and straightforward interpretations of the rules of Usenet conduct.

In examining these aspects of the data, I have adopted the terminology of Herring (1996), who uses a basic electronic message schema to analyze email discourse. According to this schema, so-called epistolary conventions form the opening and closing parts of a message. These optionally include a salutation at the beginning of a message, a complimentary close (such as "Sincerely yours") and a signature at the end of the message.

Furthermore, I have divided signatures into simple signatures, consisting of only the name of the author at the end of a message, and elaborated signatures, which also contain edited information about the author, such as an address or affiliation. Elaborated signatures are often created by users to be added automatically at the end of messages as an option in the mail system.10

Analysis of the data shows that there are striking differences between the email and the newsgroup data with respect to epistolary conventions and their relationship to quoting (see Table 1). In the email data, 54% of the quotations contain a salutation at the beginning of the included message, whereas the corresponding figure for the newsgroup data is only 2%. Similarly, 45% of the email quotations contain a complimentary close from a previous message, against 6% for the newsgroup data. However, this is partly a consequence of the fact that salutations and complimentary closes are much more common in private email (for example, only 9% of the reply messages in comp.human-factors included a salutation at the beginning of a message, compared to 67% of all the replies in the email corpus).

With this in mind, it is interesting to perform a similar analysis for elaborated signatures. These are generally more common in newsgroups (41%, compared with 29% in A-received; A herself used only a simple signature). Nevertheless, only 5% of the quotations in the newsgroup include an elaborated signature, against 13% in A-received and 19% in A-sent. This pattern confirms the contrast between the newsgroup data and the email data in that the former displays a careful, edited form of quotation, whereas the latter includes many cases where everything in a message is quoted except headers.


From a pragmatic point of view, the role of pointers preceding a quoted text is different in the two communicative contexts studied. In private email, pointers are usually unnecessary if there are only a few receivers and it is apparent who wrote the original message. They also contribute to the impersonal aspect of quoting, lending the reply a computer-generated character. In contrast, conversations in newsgroups can be difficult to follow if pointers are not used to identify which of the previous messages in a thread is being addressed in a response.

In the data, it emerged that newsgroup participants used pointers in a majority of the messages containing a quotation (79.6%). Surprisingly, the corresponding frequency in A's email data was as high as 64.9%. This appears to be a consequence of the support given to quoting in some mail systems, to the effect that pointers are part of the default setting. In fact, this was true of A's email system (Pine), which explains the fact that all of her own quoted replies contained a pointer, whereas only about one-third of the quoted replies she received contained one.11

The use of pointers may have peculiar consequences in certain cases. In embedded quotations, for example, there may be two or three pointer lines immediately following one another, making it difficult for readers to grasp the structure of the dialogue. Thus, the last message in the dialogue of Example 1 is a composite reply with the structure illustrated in Figure 2, where the boxes signify levels of embedding. Only the initial part (abbreviated as "text by B.") contains original text, wheras the rest of the message contains embedded layers of quotations including all epistolary conventions (abbreviated as "greetings" in the figure), which makes it both long and highly redundant.

Figure 2. Schematic structure of a message in A-received with three levels of quotation

An unusual case of the use of pointers is shown in Example 6, where the dialogue metaphor of quoting is exploited in order to provide a clarification to one's own previous message.

Example 6

Another case of unusual effect of pointers in the newsgroup data was when a message was posted by a person with the email address "nobody" and the reply was introduced (perhaps jokingly) with the pointer line "nobody wrote...".


As noted in the beginning of this article, quoting is a strategy particularly suited to the need for contextualization in multi-threaded discussions. The analysis of the data confirms the expectation, derived from the difference in the character of communication in the two settings, that the strategy of quoting would be most frequent in the newsgroup data. Against this background it is surprising that both A herself and the individuals corresponding by email with her use quoting frequently, and that they often include the entire message in a simple reply. This is not in line with the results of the Usenet survey (Severinson Eklundh & Macdonald, 1994), which showed that many respondents were strongly against excessive quoting, especially in private email communication. Clearly, the participants in the present study do not adhere to network etiquette with respect to quoting and do not seem to be inhibited by the fact that messages may contain large amounts of unnecessary text.

It seems, therefore, that the social rules of conduct for network communication, and especially the "Don't waste bandwidth" principle, are not given high priority in private communication within this user community. A possible explanation is that many mail systems require extra actions (and thereby knowledge on the part of the user) to quote selectively. Furthermore, the speed of the communication is often very high in private email, which may make it easier to include the whole message rather than edit it.

In spite of these results, the presence of a "Gricean conflict" with respect to quoting became evident in a short interview conducted with A about her quoting habits. Generally, she characterized quoting as a way to provide a context or important information relevant to the current message. Her use of it had gradually increased in order to avoid misunderstandings and unnecessary load for the receiver. It was evident that she gave higher priority to clarity than to brevity in her email communication. However, she also mentioned that excessive use of quotations should be avoided, and this was again motivated by a concern to avoid burdening the receiver.

The two sets of data exhibited many differences in the ways quoting was used. One was the place of the quotation in relation to the text of the response, where the newsgroup data mainly adhered to the dialogue model, with the quotation before the response or interspersed with it. In the email data, it was most common that the response preceded the quotation. Despite this difference, both sets of data typically exhibit a conversational form of response, often containing elliptic beginnings and anaphoric elements. Only in a few cases does a "document attachment model" of quoting seem relevant, e.g., when a quotation attached after a response was referred to explicitly or with a paraphrase by the writer.

The Effects of Quoting on Conversational Structure in CMC

The discussion so far has mainly focused on the characteristics of a single message. However, quoting strategies also affect a computer-mediated dialogue in a global sense. In this section, I discuss the structural properties of quoting and its role in linking turns of a computer-mediated conversation.

Quoting as a Mode of Linking in Conversation

As mentioned earlier, quoting contributes to the coherence of a computer-mediated conversation by providing an explicit link between messages (see also Herring, 1999). In particular, it strengthens the conditional relevance12 of a reply, and establishes its relationship to the original message. This is a crucial feature in delayed (asynchronous) group communication, where there may be several simultaneously ongoing dialogues on the same or different topics.

For the sender, quoting facilitates direct reply within a conversational context without paraphrasing relevant parts of the preceding discourse (Severinson Eklundh & Macdonald, 1994). The reply may therefore take advantage of the context in a way similar to spoken conversation, as shown by the abundance of ellipsis and anaphoric expressions in both corpora of reply messages in the present study. The conversational nature of the resulting interactions is particularly prominent in messages where selected quotations and responses are interspersed, lending the character of a completed dialogue to a single message (creating an "illusion of adjacency" as formulated by Herring, 1999).

In Reed (2001), an ethnomethodological account is given of the conversational structure of newsgroup interaction, showing how a conversation is ordered through participants' quoting practices. The analysis focuses specifically on the dialogic pattern exhibited by the prototypical newsgroup message with interspersed quotes (called a "turn-taking unit" by Reed) which embodies part of the history of the preceding conversation. Through a set of examples, it is shown how the "sequential integrity" of a turn-taking unit is maintained as a result of participants' actions, sometimes going against the system's default positioning of a response in relation to quoted text (p. 8).

As a conversational strategy, quoting can be seen as a way of constraining the response and thereby the continued dialogue. By selecting a part of the message to which a response is provided, the sender simultaneously excludes a number of other possible interpretations of the message and focuses explicitly on the quoted parts as salient or noticeable. In contrast, the participants in an ordinary conversation can respond to an utterance in a "loose" way, without explicitly specifying which aspects of a message are being addressed in the answer.13 This potential ambiguity gives the interlocutors greater freedom to negotiate meaning, compared to the referential explicitness required by a quotation-response dialogue. In fact, our Usenet survey showed that pointing out a certain part of a message by quoting it is often avoided in sensitive discussions.

Quoting may be pragmatically contrasted with threading as a coherence mechanism in a computer-mediated dialogue. Threading has the general role of preserving the Reply-to relationship between messages with a common subject, and thereby to contextualize a reply message. This relationship may be emphasized in different ways depending on the system design. Quoting, in contrast, serves to contextualize a response as an individual action (not the entire message which is being replied to) and also to select which parts of the message are to be responded to. Thus, one might say that quoting technically presupposes a threading mechanism (since the system inserts a copy of the message only when the Reply command is selected), but it goes beyond the message structure to provide a local sequential organisation of the dialogue (cf. Reed, 2001).

The selective aspect of quoting holds not only with respect to the relationship between two successive turns, but also with respect to how a dialogue develops in a global sense. In spoken conversations, it is possible to connect a response implicitly to several past utterances, or to the whole topic under discussion, simply by phrasing the message in a way that has a bearing on several previous turns. In CMC discussions it is generally difficult to attach to more than one utterance, at least formally, since the system makes a link to a single previous message when the Reply command is used. Quoting makes this linking to a single message more explicit by displaying the actual text referred to in the Reply.

Making a reference to several previous contributions, e.g., in newsgroups, may of course be done by mentioning them explicitly. However, results from our Usenet survey showed that many users avoided extensive paraphrasing and preferred direct response.14 To some extent the use of embedded quotations makes room for "multiple voices" in a message, though this requires careful use and sometimes editing in order to produce a clear and readable representation of the discussion.

Many forms of conversational linking are difficult to accomplish without paraphrasing. One example is provided in the newsgroup data in our corpus, where a participant needed to refer both to a previous contribution and to the first message in the thread in order to motivate her own stance on a particular topic (Example 7).

Example 7

In reviewing the discussion, the writer first quotes from and responds to a previous message. She then inserts a "manual" quote from the initial message in the thread, and gives a paraphrase of subsequent comments, followed by a declaration of her own view. This kind of solution, combining quotation with meta-comments to summarize a conversation, illustrates the participants' needs for more advanced intertextual reference mechanisms.

Sometimes participants need to connect not to the posting that they have just read, but to an earlier message, perhaps referenced in the current one. In such cases it is usually simpler to attach one's response to the currently displayed message by using the Reply command than to locate the previous message in the thread. The presence of a relevant quotation in the current message may facilitate this. Thus, in two cases in the newsgroup data, a participant simply quotes the quotation of the previous speaker instead of going to the original message and quoting directly from it.

In other cases, this use of the Reply command may result in a misleading dialogue structure. In Example 8, a newsgroup participant quotes and responds to a current reply instead of the question that he wishes to address.15 The resulting dialogue has the structure <Response1 - Response 2>, where both responses address the same (unquoted) original message. Although a similar "loose" way of answering would be acceptable in ordinary conversation, expectations connected with the regular quote-response structure may here give the reader the false appearance of a minimal adjacency pair.

Example 8

The use of pointers in newsgroup communication further accentuates the discrepancies which may occur due to this form of replying. For example, by automatically pointing out the source of a quotation, the system assumes that this message is the one being responded to. However, in at least one case in the newsgroup data, an author quotes a message that already contains quotations and responds to the second quotation's source, although the attached pointer mentions the previous one; a seeming contradiction which may make it more difficult for participants to detect the intended dialogue structure.

Playful and Deviant Applications of Quoting

Like other aspects of the computer medium, quoting sometimes invites playful use in which participants extend the limits that it sets for a conversation. In other cases its application leads to errors or mistakes. Such unexpected cases often have interesting conversational consequences.

Some examples in this category have already been mentioned. They involve an unusual source of a quotation, such as the case of "quoting oneself" when someone feels the need to explain his or her own utterance, or "quoting nobody" as a result of an unusual email address. The use of pointers accentuates the playful aspect of these cases.

The quote-response pattern may also be used in new ways to simulate a dialogue. A case which is not uncommon is when someone quotes a previous speaker's signature or greeting. If this is followed by a reciprocation, it can be seen as a social way of ending a conversation (taking the form of a "leave-taking" adjacency pair). However, sometimes a quoted signature is left at the end of a message with no comment, apparently because the writer did not take time to remove it. This could create misunderstandings concerning who is the author of the message.

When users edit their reply, mistakes may easily occur. One type of error that occurred once in the corpus is that an author mistakenly wrote the answer within the quote sign, so that it looked like part of the quote (Example 9). This situation – conflating the contributions of different speakers – has no direct counterpart in ordinary, spoken conversation.

Example 9

Consequences of the Regular Quote-response Pattern

Apart from linking contributions in computer-mediated discourse, quoting restricts the pragmatic choices for shaping the response. By constraining how answers are composed and providing slots for a topic-oriented response, quoting imposes its own structure on the reply message.

This structure helps the sender to take advantage of the discourse context, since the reader can use the presence of the quoted text to resolve these references. But also, the presence of quotation prescribes an order for the answer, which may otherwise be a problem for users when there are several topics or issues that need to be addressed. In Severinson Eklundh (1986), examples are presented suggesting that there are conflicting strategies for answering multi-topic messages in CMC conversations, corresponding to more "spoken" as opposed to more "written" norms for using the medium.

This rigidity of the quote-response structure may create a certain striving for completeness and symmetry. This is most evident in cases where a long message is quoted and responded to in chunks. Participants sometimes attempt to quote and respond in a symmetrical way corresponding to the parts of the original message, even if this is not directly motivated by the content of the intended answer. One example is from A's mailbox data, where a participant inserts a feedback signal ("That's interesting") in the middle of a long quoted passage, while interspersing longer answers to other parts of the same message. One may doubt that the author would have bothered to comment on this if the quote were not there to provide a context. Giving explicit, independent feedback is an uncommon strategy in email due to the time delay and the "cost" of sending a new message (Severinson Eklundh, 1986).16 In this case, one might say that quoting allows a participant to give feedback without using extra bandwidth in the form of an independent message.

In many cases, the tendency for symmetry or completeness in the interspersed quotation-response creates "left-over" quotations in the reply. In Example 10, it seems that the author of the response wants to avoid deleting only the last part of the original message, even though he has nothing particular to add. This yields a certain ambiguity as to whether he intends to comment on this passage or not.

Example 10

In summary, the quote-response structure makes answering easier by providing a schema for the reply. It also appears to lead to a concern for completeness in the way answers are composed. Furthermore, the structure invites new uses and experimentation, as well as leading to occasional errors which may cause misunderstandings about the structure of a dialogue.


In this article, I have shown that contextual quotation is a pragmatic and interactive device that is frequently used both in the email and in the newsgroup corpus. It serves both as a global mode of linking between successive discourse contributions and as a structuring mechanism for the individual reply message.

Generally, quoting creates a fixed, explicit linking between successive discourse contributions rather than loose connections of the kind typical of ordinary conversation. In its selective form, it allows an interlocutor to specify explicitly what aspect of an utterance he or she is taking a stand in response to. This may be part of the reason why quoting sometimes is perceived as impolite. It draws attention to the exact form of the message being responded to or brings a special part of it into focus, and it may sometimes be perceived as doing this as part of a negative stance (cf. the findings of Hodsdon-Champeon, this issue).

In this study, private email users were found to use quoting less frequently, and less selectively, than newsgroup participants. Table 2 summarizes and contrasts the typical pattern found in these two settings. Whereas the newsgroup data exhibited carefully edited, topic-oriented quotation patterns, the email authors often included a whole message when they quoted it. A typical reply message therefore was lengthy and redundant, including pointers, salutations and signatures. This suggests that the newsgroup discussions are more regulated by "netiquette" rules that oblige the participants to adhere to Gricean conversational maxims for the purposes of clarity, relevance, and economy. In particular, violating the principle "do not waste bandwidth" may lead to negative sanctions in Usenet discussions (McLaughlin et al., 1995).

Table 2. Summary of quoting patterns in two settings

In general, quoting can be seen as playing a more constraining role for ongoing dialogue in newsgroup discussions than in private email, as exhibited in the data analyzed here. Because newsgroup participants regularly edit the quoted message and select what to respond to, they can also determine to a greater extent what the following conversation will focus on. The point-by-point quote-response pattern creates an impression of a dialogue which is completely on the sender's terms. Therefore, quoting acquires a more strategic, argumentative role in newsgroups as compared to the email data, where it serves mainly to help the reader contextualize the reply.

The use of quoting depends to a significant extent on the design of the mail system. For example, certain mail systems require the user to insert a whole message first and then decide what to quote by removing text material, whereas other systems allow the user to select the material to be quoted directly. Thus details of the interface may play an important role in determining what quotation patterns emerge. For example, if a user is able to select a passage by marking it with the mouse, he or she can just as well mark a sentence or other relevant linguistic unit, whereas other systems may enforce less flexible, row-wise quoting that does not correspond to the way the original message and the response are actually linked.

The design of the mail system used by A in this study (Pine) is likely to have influenced her quoting strategies in several respects. One example is her consistent placement of the quotation below the original response text. With respect to using quoting or not, however, the system gives comparable support to both options,17 and therefore her choices in this respect must be explained in other ways.

Technical factors may be intertwined with social ones in shaping participants' strategies. Notably, the participants in the Usenet survey whose systems included messages as a default in the reply command reported using quotation as a linking device twice as often as other users. At the same time, as many as 25% of all participants used quoting regularly in spite of there being no support for it in their system (e.g., by copying and pasting material from a previous message). This constitutes evidence that quoting is part of an established norm system for network communication (cf. McLaughlin et al., 1995) operating on Usenet as well as in many other CMC settings.

It is interesting to compare the results of this study with an early study of dialogues in the Swedish COM system, which was the subject of my dissertation work (Severinson Eklundh, 1986). In the email messages collected from COM, there were no instances of quoting. This can be explained partly by the fact that the system did not provide support for inclusion of messages. But also, the COM community was a homogeneous social environment, and few if any users were immediately connected to a larger network. The conversational style was therefore not directly influenced by norms of international network communication with respect to the demand for relevance and information density. In spite of the absence of quotes in the COM message corpus, the dialogues had a strongly conversational character, with context-dependent, short and informal responses as the prototypical case. This style was encouraged by the design of the system, which contained commands for reviewing the current dialogue while reading or writing.

The norms for network communication tend to change over time, as a result of both technology development and users’ appropriation of the technology in new social settings. The data analyzed in the present study were collected in 1996. Since then, the use of the Internet has continued to grow and new forms of communication have emerged. At present, Usenet news is less important than before as a discussion medium18 (see, e.g., Wikipedia, August 26, 2010), whereas mailing lists, web boards, and blogs have increased in importance. The use of email continues to proliferate, however, and the computer has taken over traditional paper communication in many contexts. Quoting entire messages in email seems to have become more common in many work-related and academic contexts (see, e.g., Duchenaut & Bellotti, 2003), and well-known email clients enforce this habit through their default settings.

In conclusion, how computer users employ a mechanism like quoting is affected by the social environment, as well as by technical factors in CMC systems. In many cases, however, users do not act in the way the system or the environment expects them to, or they simply do not take the time to use the technology in the most rational or appropriate manner. The data analyzed here include many examples of deviant and playful instances of quoting, which seem to develop from the interplay between system, individual, and social factors. In this as in other respects, the means of communication offered by computer-based media are constantly reshaped by their users.


  1. Usenet is a vast communication network system that allows people to exchange messages and engage in discussions on various subjects. It was established around 1980 and is still used, although its importance has diminished with the development of the web and new forms of net-based communication. The articles that users post to Usenet are organized into topical categories called newsgroups. At present, Google Groups hosts an archive of Usenet posts dating back to May 1981 (Wikipedia, August 26, 2010).

  2. Tannen (1989) discusses numerous examples of how repetition on the word, phrase, and sentence levels is used to create coherence and interpersonal involvement in spoken conversation.

  3. Du Bartell (1995) has suggested the term "adhesion" for what I call contextual quotation here.

  4. The names appearing in the examples from the collected data have been changed or omitted. The messages are presented in a simplified way, henceforth without the automatically generated header information about sender, receiver, date, etc. The examples from A's mailbox have been translated from Swedish into English.

  5. Negative politeness is behavior intended to protect participants' negative face, i.e., "the want of every 'competent adult member' that his actions be unimpeded by others" (Brown & Levinson, 1987, p. 62).

  6. A used Pine, an email program well-known at the time, which runs in the Unix environment. Pine does not have a graphic interface, but uses text-based menus to present options to the user.

  7. It was not possible to determine with absolute certainty how many messages were produced during the period in question, due to some irregularities in the local access to the news service.

  8. Although I point to and discuss a number of ways in which the two types of data differ, I do not perform any advanced statistical analysis to compare them rigorously. The reason is both the limited size of the corpus and the fact that there are so many unique contextual characteristics of these two forms of CMC.

  9. Korenman and Wyatt (1996) analyzed a number of response strategies on a Women's Studies discussion list, some of them related to the ones dealt with here. On the basis of their findings, they argued that asynchronous CMC is conceptualized as "pseudo-oral" by users.

  10. Some users develop personal signatures, containing simple graphics, proverbs, etc. In my analysis, I used the criterion that either the address (physical or email) or the affiliation of the author should appear in an elaborated signature. The reason is simply that it is not possible to state with certainty if a piece of text following a name is intended to be part of the signature.

  11. It should be noted that a pointer line as well as any other part of a message may always be removed by the user while editing the reply. Keeping the pointer can therefore be seen as a voluntary choice, although it is difficult to assess the extent to which a user is aware of this possibility. In the data, there are only two cases of a user who makes use of both strategies (quoting with and without a pointer).

  12. Conditional relevance is a central concept in the research tradition of conversational analysis. It refers to the relation of dependence between the two parts of an adjacency pair, expressing the criterion that given the first part of an adjacency pair, the second part is relevant and expectable (Schegloff, 1972). The concept can be applied independently of some of the criticisms that have been directed towards the adjacency pair concept (see Severinson Eklundh, 1983). Levinson (1983) discusses the notion of conditional relevance within the framework of a more general "preference organization" of conversation.

  13. Rather than providing a direct response, a participant can connect to a conversation in a flexible way by using referents or concepts associated with a previous turn. The resulting topical structure emerges informally, partly established through nonverbal cues. Levinson (1983) gives examples to illustrate that topical coherence in this regard cannot be described by rules; rather, it is something constructed across turns by the collaboration of participants (p. 315).

  14. Similar results were found in an observational study of a discussion forum by Korenman and Wyatt (1996), who found that participants very often posted contributions with an "inadequate summary of a previous message." In fact, this was the most frequent violation of the guidelines distributed to all new users.

  15. In this example, a colon is used as a quotation marker.

  16. In Severinson Eklundh (1986), a body of data from the COM system was analyzed with respect to the structural properties of a computer-mediated dialogue. The absence of "backchannel" feedback and minimal evaluative responses was found to be a characteristic property of the data, which was traced to the general conditions for asynchronous CMC. It was argued that this lack of feedback may lead to an uncertainty as regards the state and outcome of a dialogue (creating "dangling conversations," cf. Black, Levin, Mehan & Quinn, 1983).

  17. In Pine, the user is prompted with a question "Include original message in Reply?" when issuing the Reply command. The answers Y(es) and N(o) are equally easy to provide. Just pressing the Return key will amount to the default choice "No." Therefore, the interface design of Pine cannot be seen as encouraging a consistent choice of quoting.

  18. Smith (1999) provides a thorough description of quantitative aspects of Usenet communication. His presentation includes posting frequencies in various newsgroups, covering the period when the data for the present study were collected.


Black, S. D., Levin, J. A., Mehan, H., & Quinn, C. N. (1983). Real and non-real time interaction: Unraveling multiple threads of discourse. Discourse Processes, 6, 59-75.

Brown, P., & Levinson, S. C. (1987). Politeness: Some universals in language usage. Cambridge: Cambridge University Press.

Du Bartell, D. (1995). The functions of adhesion in computer-mediated discourse. Paper presented at a workshop on Computer-Mediated Discourse Analysis, Georgetown University Roundtable in Linguistics, March 1995.

Duchenaut, N., & Bellotti, V. (2003) Ceci n’est pas un objet? Talking about objects in email. Human-Computer Interaction, 18, 85-110.

Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and Semantics 3: Speech Acts. New York: Academic Press.

Herring, S. C. (1994). Politeness in computer culture: Why women thank and men flame. In M. Bucholtz (Ed.), Cultural Performances: Proceedings of the Berkeley Women and Language Conference (pp. 278-294). Berkeley: Berkeley Women and Language Group.

Herring, S. C. (1996). Two variants of an electronic message schema. In S. C. Herring (Ed.), Computer-Mediated Communication: Linguistic, Social and Cross-Cultural Perspectives (pp. 81-108). Amsterdam and Philadelphia: John Benjamins.

Herring, S. C. (1999). Interactional coherence in CMC. Journal of Computer-Mediated Communication, 4 (4). http://jcmc.indiana.edu/vol4/issue4

Hodsdon-Champeon, C. B. (This issue.) Conversations within conversations: Intertextuality in racially antagonistic dialogue on Usenet.

Korenman, J. & Wyatt, N. (1996). Group dynamics on an email forum. In S. C. Herring (Ed.), Computer-Mediated Communication: Linguistic, Social and Cross-Cultural Perspectives. Amsterdam: John Benjamins.

Levinson, S. (1983). Pragmatics. Cambridge: Cambridge University Press.

McLaughlin, M. L. (1984). Conversation: How talk is organized. Beverley Hills: Sage Publications.

McLaughlin, M. L., Osborne, K. K., & Smith, C. B. (1995). Standards of conduct on Usenet. In S. G. Jones (Ed.), CyberSociety. Thousand Oaks: Sage Publications.

Reed, D. (2001). 'Making conversation': Sequential integrity and the local management of interaction on Internet newsgroups. Proceedings of the 34th Annual HawaFi International Conference on System Sciences. Los Alamitos: IEEE Press.

Schegloff, E. A. (1972). Sequencing in conversational openings. In J. J. Gumperz & D. H. Hymes (Eds.), Directions in Sociolinguistics. NY: Holt, Rinehart, & Winston.

Severinson Eklundh, K. (1983). The notion of language game: A natural unit of dialogue and discourse. Studies in Communication, 5. Linköping, Sweden: Department of Communication Studies, University of Linköping.

Severinson Eklundh, K. (1986). Dialogue processes in computer-mediated communication: the study of letters in the COM system. (Diss.) Linköping Studies in Arts and Science 6. Linköping, Sweden: University of Linköping.

Severinson Eklundh, K., & Macdonald, J. (1994). The use of quoting to preserve context in electronic mail dialogues. IEEE Transactions on Professional Communication, 37 (4), 197-202.

Smith, M. A. (1999) Invisible crowds in cyberspace: mapping the social structure of the Usenet. In M. A. Smith & P. Kollock, Communities in Cyberspace. New York: Routledge.

Sudweeks, F., & Rafaeli, S. (1996). How do you get a hundred strangers to agree? Computer mediated communication and collaboration. In T. M. Harrison & T. D. Stephen (Eds.), Computer networking and scholarly communication in twenty-first century university. Albany, NY: SUNY Press.

Tannen, D. (1989). Talking voices. Repetition, dialogue, and imagery in conversational discourse. Cambridge: Cambridge University Press.

Webster's (1993). Random House Webster's Electronic Dictionary and Thesaurus, College Edition, Version 1. Working Software, Inc.

Wikipedia. (2010). Usenet. http://en.wikipedia.org/wiki/Usenet, last modified August 26, 2010.

Biographical Note

Kerstin Severinson Eklundh (kseF "at" csc "dot" kth "dot"se ) is professor emerita in Human-Computer Interaction at the Royal Institute of Technology (KTH) in Stockholm. Starting with her Ph.D. in the 1980s, she has a long-standing interest in computer-mediated dialogue processes. Her recent research topics include computer-supported cooperative work, writing tools, and human-robot interaction.


Any party may pass on this Work by electronic means and make it available for download under the terms and conditions of the Digital Peer Publishing License. The text of the license may be accessed and retrieved at http://www.dipp.nrw.de/lizenzen/dppl/dppl/DPPL_v2_en_06-2004.html.